Two AIs walk into a negotiation. Only one walks out richer.

Nobody became a developer to write documentation. But the docs still need to get written — PRDs, README updates, architecture decisions, onboarding guides.

Wispr Flow lets you talk through it instead. Speak naturally about what the code does, how it works, and why you built it that way. Flow formats everything into clean, professional text you can paste into Notion, Confluence, or GitHub.

Used by engineering teams at OpenAI, Vercel, and Clay. 89% of messages sent with zero edits. Works system-wide on Mac, Windows, and iPhone.

Try Wispr Flow free

A private marketplace. Like Craigslist, but with one twist.

No human did the negotiating. Every employee handed the keys to an AI agent and told it: go buy and sell things on my behalf.

69 employees. 500+ items. One week.

By the end, the agents had closed 186 deals worth just over $4,000 in real money. Snowboards, plants, board games, a wooden bowl. One agent bought 19 ping-pong balls as a gift for Claude itself, which it described as "perfectly spherical orbs of possibility."

Cute story. But that's not the part that should make you sit up.

Here's the part that should.

Anthropic quietly split participants into two groups. Half were represented by Claude Opus, the strongest model. Half by Claude Haiku, the lightweight one. Same instructions. Same marketplace. Same rules.

The Opus agents won. Every time.

Sellers using Opus earned $2.68 more per item, on average. Buyers using Opus saved $2.45 per item. Opus users closed about 2 more deals overall. When an Opus seller met a Haiku buyer, the average sale price was $24.18, compared to $18.63 in Opus-versus-Opus deals.

The weaker model didn't just lose. It got systematically out-negotiated, deal after deal, without ever realising it was happening.

And here's the kicker.

The humans on the losing side rated the experience as fair.

They thought their agent did a great job. They thought the deals were reasonable. When asked afterwards to rank their bundles, 11 of the 28 people who ran with both models actually preferred the Haiku run. The one where they got worse outcomes.

They didn't know they'd lost. They didn't know there was anything to lose.

This is the future most people aren't ready for.

Not "AI will replace you." That framing is lazy and Anthropic's experiment proves it wrong. The agents didn't replace humans. They represented them. The humans set the strategy. The agents executed it.

The real shift is quieter and more uncomfortable.

In a world where AI handles the negotiation, the procurement, the vendor outreach, the pricing emails, the customer follow-ups, the quality of your AI tooling becomes a hidden tax on your business. Two competitors selling the same product can have completely different margins, completely different close rates, completely different growth curves, simply because one of them is running a sharper system underneath.

And the one losing won't see it. Because everything will feel fine. Deals will close. Customers will sign. Numbers will look reasonable.

Just slightly worse than the guy across town.

Here's the part that matters for you, today.

If you're a business owner reading this, the "should I use AI" question is already settled. You are using it. Or your competitor is. Either way, it's in your business.

The real question is whether the version you're using is the strongest one, configured properly, talking to the rest of your operation. Or whether it's a free tier, a generic prompt, a half-baked workflow you set up six months ago and forgot about.

Because Project Deal showed something most AI hype articles miss: in a head-to-head match, the better-equipped side wins quietly. The losing side doesn't get a notification. They just leave money on the table, day after day, and call it Tuesday.

Hit reply and tell me one place in your business where AI is doing the negotiating, the writing, or the deciding right now. I'm curious which workflows people actually trust it with, and which ones they don't.

Stay sharp.